I have a bunch of images, that follow this pattern.
0001.png
0004.png
0007.png
0010.png
0013.png
0016.png
0019.png
...
They're in the correct order, but FFMPEG ignores anything after the first file, because it can't find a 0002.png. How can I get it to join these files?
I tried
ffmpeg -i %04d.png out.avi
For recent versions of ffmpeg, see Gyan's answer instead
Original answer:
with ffmpeg 0.11.1 it's as easy as:
ffmpeg -f image2 -i %*.png out.avi
From the man page, in an example under "Video and Audio file format conversion":
When importing an image sequence,
-ialso supports expanding shell-like wildcard patterns (globbing) internally. To lower the chance of interfering with your actual file names and the shell's glob expansion, you are required to activate glob meta characters by prefixing them with a single%character, like infoo-%*.jpeg,foo-%?%?%?.jpegorfoo-00%[234%]%*.jpeg.
Update per comments: on recent versions you should now use -i '*.png' syntax.
glob_sequence pattern type is deprecated, but the glob pattern type is not. The answer by Mulvya and Peter Cordes works.
Commented
Sep 7, 2015 at 22:03
Pattern type 'glob_sequence' is deprecated: use pattern_type 'glob' instead
Commented
Feb 7, 2017 at 16:17
Pattern type 'glob_sequence' is deprecated...): the command given by rymo above (specifically, -i %*.png) will cause the warning. To eliminate it, use the slightly-different syntax in Peter Cordes answer (-i '*.png').
Besides the workaround in my comment, you should be able to use this syntax from the ffmpeg documentation:
ffmpeg -f image2 -pattern_type glob -i '*.png' out.avi
The quotes are important, you need ffmpeg to see the *, not have the shell expand it.
*.png in single-quotes (so '*.png') to avoid it being expanded by Unix-style shells (so, OSX and linux etc).
From http://www.ffmpeg.org/faq.html: If you have large number of pictures to rename, you can use the following command to ease the burden. The command, using the bourne shell syntax, symbolically links all files in the current directory that match *jpg to the ‘/tmp’ directory in the sequence of ‘img001.jpg’, ‘img002.jpg’ and so on.
x=1; for i in *jpg; do counter=$(printf %03d $x); ln -s "$i" /tmp/img"$counter".jpg; x=$(($x+1)); done
seqA-[0001-n].png, seqB-[0001-n].png, etc. Much easier than messing around with image2 globbing and the concat filter.
for i in *jpg with for i in /full/path/*jpg, because all the symlinks were broken
I would actually suggest the same thing Mulvya suggested. Whenever I've needed to do stuff like this when dealing with 3D animation renderings, I've renamed the files to make them sequential. It's just easier to work with that way in a lot of different software unless you need to preserve the file names for some reason. Total Commander is also a great tool for multi-renaming.
For the missing sequence part, I used 2 liner
CONCAT=$(echo $(ls *.png | sort -n -t _ -k 2) | sed -e "s/ /|/g")
ffmpeg -framerate 25 -i "concat:$CONCAT" -c:v libx264 -profile:v high -pix_fmt yuv420p output.mp4
Hope this will help someone in future
If the images are dated, exiftool can be used to rename the source files to yyyy-mm-dd_hhmm pattern. Then a bash script can create an input file for ffmpeg.
Here is a script that shows the idea: https://github.com/leobard/timelapse-prepare-and-run-ffmpeg.sh/blob/main/timelapse-prepare-and-run-ffmpeg.sh
